SlideShare uma empresa Scribd logo
1 de 23
Design Choices for Cluster-
Randomised Trials
Alan Girling
University of Birmingham, UK
A.J.Girling@bham.ac.uk
CLAHRC Scientific Advisory
Group Birmingham, June 2015
… a Statistical viewpoint
• Ethical and Logistical concerns are suspended
• RCTs and Parallel Cluster trials are well-
understood; ‘Stepped Designs’ less so.
Two questions:
1. How does the Stepped Wedge Design perform?
…but The Genie is out of the bottle!
2. What about alternative stepped designs?
Does it have to be a “Wedge”?
Some Alternative Designs:
8 months 4 groups of Clusters (“Arms”)
Arms
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Months 
Arms
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Months 
Assumptions
• Study of fixed duration (8 months)
• Constant recruitment rate in each cluster
• Continuous Outcome
– Additive treatment effect
– Cross-sectional observations with constant ICC = 
• Cross-over in one direction only (i.e. Treated to Control prohibited)
• The analysis allows for a secular trend (“time effect”)
– This has been questioned; but if time effects are ignored, the ‘best’
statistical design involves simple before-and-after studies in each cluster
(i.e. not good at all!)
Goal: To compare statistical performance of different designs
under these assumptions, especially the Stepped-Wedge
1. ‘Precision Factor Plots’ for
Comparing Designs
Clusters
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Months 
1. Simple Parallel Design
Two groups of clusters.
Treatment implemented in
one group only, in month 1.
Treatment Effect Estimate =
(Mean difference between two groups over all months)
Two Simple Candidate Designs
Clusters
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Months 
1. Parallel Study with (multiple) baseline controls
“Controlled Before-and-After Design”
Two groups of clusters.
Treatment implemented in
one group only, in month 5.
Treatment Effect Estimate =
(Mean difference between two groups in months 5 – 8)
minus
r x (Mean difference between two groups in months 1 – 4)
(r is a correlation coefficient “derived from the ICC”)
Performance measured by Precision of the effect estimate:
Precision = 1/(Sampling Variance) = 1/(Standard Error)2
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
 













R
N
4
1
2
1
14 2
  
 R
N







1
14 2

Precision =
Where R is the “Cluster-Mean Correlation”
… and is the same for both designs  

11 

m
m
R
(m = number of observations per cluster,  = ICC)
CBA Parallel
Relative Efficiency of designs – by comparing straight
lines on a Precision-Factor plot
CBA design is better (“more efficient”) if R > 2/3.
Otherwise the Parallel design is better.
Parallel
CBA
2. The Cluster-Mean Correlation
(CMC)
Cluster-Mean Correlation (CMC)
• Relative efficiency of different designs depends on
cluster-size (m) and ICC (), but only through the CMC (R)
• The CMC (R) =
proportion of the variance of the average
observation in a cluster that is attributable to
differences between clusters
• (The ICC () =
proportion of variance of a single observation
attributable to differences between clusters)
• CMC can be large (close to 1) even if ICC is small
  



m
m
m
m
R




111
Cluster-Mean Correlation: relation with cluster size
• CMC can be large (close to 1) even if ICC is small
• So CBA can be more efficient than a parallel design for
reasonable values of the ICC, if the clusters are large enough




 11
m
R
R
3. Some other designs
1 1 1 1 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Controlled Before & After Before & After + Parallel
R
4
1
2
1
 R
2
1
4
3

0 1 1 1 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
0 0 0 0 0 0 0 1
R
32
9
16
9

0 1 1 1 1 1 1 1
0 0 0 1 1 1 1 1
0 0 0 0 0 1 1 1
0 0 0 0 0 0 0 1
R
16
5
8
5

+ Baseline & Full Implementation Stepped Wedge
Precision
Factors
• Stepped-Wedge is best when the CMC is close to 1
• Parallel Design is best when the CMC is close to 0 – but risky if ICC
is uncertain
• Before & After/Parallel mixture is a possible compromise
Parallel
Stepped-Wedge
B & A/Parallel
4. Is there a ‘Best’ Design?
• If Cross-over from Treatment to Control is permitted,
the Bi-directional Cross-Over (BCO) design has Precision
Factor = 1 at every R, better than any other design.
0 0 0 0 1 1 1 1
0 0 0 0 1 1 1 1
1 1 1 1 0 0 0 0
1 1 1 1 0 0 0 0
R01
Precision Factor =
• But this design is usually not feasible!
• If ‘reverse cross-over’ is disallowed, the Precision Factor
cannot exceed 2
3
1
1 RR 
1 1 1 1 1 1 1 1
0 0 1 1 1 1 1 1
0 0 0 0 0 0 1 1
0 0 0 0 0 0 0 0
• Available design performance limited by choosing only 4 groups of
clusters
Some (nearly) ‘Best’ designs with 4 groups of clusters
Stepped-
Wedge
Parallel
Region
Prohibited by
Irreversibility
of Intervention
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
0 1 1 1 1 1 1 1
0 0 0 1 1 1 1 1
0 0 0 0 0 1 1 1
0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Some (nearly) ‘Best’ designs with 8 groups of clusters
Stepped-
Wedge
Parallel
‘Best’ Design for large studies: This is a mixture
of Parallel and Stepped-Wedge clusters.
100R%
Stepped-Wedge
Clusters
100(1 – R)%
Parallel
Clusters
• Includes Parallel and Stepped-Wedge designs as special cases
• When R is close to 1 the Stepped-Wedge is the best possible
design
5. Conclusions
• Efficiency of different cluster designs depends on the
ICC () and the Cluster-size (m) but only through the
CMC (R).
• Precision Factor plots are useful for comparing designs
– Comparisons are linear in R
• The Stepped-Wedge Design is most advantageous
when R is close to 1.
– In studies with large clusters this can arise even if the ICC is
relatively small.
• The theoretically ‘best’ design choice is sensitive to R,
and combines Parallel with Stepped-Wedge clusters
Limitations
• Continuous data, simple mixed model
– Natural starting point – exact results are possible
– ?Applies to Binary observations through large-
sample approximations
• Extension to cohort designs through nested
subject effects is straightforward and gives
essentially the same answers

Mais conteúdo relacionado

Destaque

Welcome and introduction...The year that was - Prof Richard Lilford
Welcome and introduction...The year that was - Prof Richard LilfordWelcome and introduction...The year that was - Prof Richard Lilford
Welcome and introduction...The year that was - Prof Richard LilfordNIHR CLAHRC West Midlands
 
Publication bias in service delivery research: a critique - John Ovretveit
Publication bias in service delivery research: a critique - John OvretveitPublication bias in service delivery research: a critique - John Ovretveit
Publication bias in service delivery research: a critique - John OvretveitNIHR CLAHRC West Midlands
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_af
Clahrc ps cmeeting_21st_sept2015_spacer_project_afClahrc ps cmeeting_21st_sept2015_spacer_project_af
Clahrc ps cmeeting_21st_sept2015_spacer_project_afNIHR CLAHRC West Midlands
 
Modelling causal pathways in health services part 2 - Sam Watson
Modelling causal pathways in health services part 2 - Sam WatsonModelling causal pathways in health services part 2 - Sam Watson
Modelling causal pathways in health services part 2 - Sam WatsonNIHR CLAHRC West Midlands
 
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...NIHR CLAHRC West Midlands
 
Merhaba demek
Merhaba demekMerhaba demek
Merhaba demekendaze41
 

Destaque (8)

Modelling and Estimation
Modelling and EstimationModelling and Estimation
Modelling and Estimation
 
Welcome and introduction...The year that was - Prof Richard Lilford
Welcome and introduction...The year that was - Prof Richard LilfordWelcome and introduction...The year that was - Prof Richard Lilford
Welcome and introduction...The year that was - Prof Richard Lilford
 
Publication bias in service delivery research: a critique - John Ovretveit
Publication bias in service delivery research: a critique - John OvretveitPublication bias in service delivery research: a critique - John Ovretveit
Publication bias in service delivery research: a critique - John Ovretveit
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_af
Clahrc ps cmeeting_21st_sept2015_spacer_project_afClahrc ps cmeeting_21st_sept2015_spacer_project_af
Clahrc ps cmeeting_21st_sept2015_spacer_project_af
 
Modelling causal pathways in health services part 2 - Sam Watson
Modelling causal pathways in health services part 2 - Sam WatsonModelling causal pathways in health services part 2 - Sam Watson
Modelling causal pathways in health services part 2 - Sam Watson
 
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...
Observations and Reflections of my visit to Rolls-Royce in Derby - Nathalie M...
 
Oxford home study prospectus
Oxford home study prospectus Oxford home study prospectus
Oxford home study prospectus
 
Merhaba demek
Merhaba demekMerhaba demek
Merhaba demek
 

Semelhante a Design choices for cluster randomised trials - Alan Girling

The stepped wedge cluster randomised trial workshop: session 4
The stepped wedge cluster randomised trial workshop: session 4The stepped wedge cluster randomised trial workshop: session 4
The stepped wedge cluster randomised trial workshop: session 4Karla hemming
 
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla Hemming
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla HemmingDog-legs, Ladders and a CONSORT statement - Alan Girling and Karla Hemming
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla HemmingNIHR CLAHRC West Midlands
 
Design of Engineering Experiments Part 5
Design of Engineering Experiments Part 5Design of Engineering Experiments Part 5
Design of Engineering Experiments Part 5Stats Statswork
 
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...Borhan Kazimipour
 
Normal Distribution
Normal DistributionNormal Distribution
Normal DistributionCIToolkit
 
Numerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsNumerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsPhD ISG, Sapienza University of Rome
 
Central Composite Design
Central Composite DesignCentral Composite Design
Central Composite DesignRuchir Shah
 
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...Borhan Kazimipour
 
Revisiting the Calibration of Modern Neural Networks
Revisiting the Calibration of Modern Neural NetworksRevisiting the Calibration of Modern Neural Networks
Revisiting the Calibration of Modern Neural NetworksSungchul Kim
 
ARIMA Model for analysis of time series data.ppt
ARIMA Model for analysis of time series data.pptARIMA Model for analysis of time series data.ppt
ARIMA Model for analysis of time series data.pptREFOTDEBuea
 
Bounded Model Checking
Bounded Model CheckingBounded Model Checking
Bounded Model CheckingIlham Amezzane
 
Reporting.ppt
Reporting.pptReporting.ppt
Reporting.pptasodiatel
 
Bioequivalence experimental study design By Vishnu Datta M
Bioequivalence experimental study design By Vishnu Datta MBioequivalence experimental study design By Vishnu Datta M
Bioequivalence experimental study design By Vishnu Datta MVishnu Datta Maremanda
 

Semelhante a Design choices for cluster randomised trials - Alan Girling (20)

The stepped wedge cluster randomised trial workshop: session 4
The stepped wedge cluster randomised trial workshop: session 4The stepped wedge cluster randomised trial workshop: session 4
The stepped wedge cluster randomised trial workshop: session 4
 
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla Hemming
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla HemmingDog-legs, Ladders and a CONSORT statement - Alan Girling and Karla Hemming
Dog-legs, Ladders and a CONSORT statement - Alan Girling and Karla Hemming
 
Design of Engineering Experiments Part 5
Design of Engineering Experiments Part 5Design of Engineering Experiments Part 5
Design of Engineering Experiments Part 5
 
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
 
Normal Distribution
Normal DistributionNormal Distribution
Normal Distribution
 
Numerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsNumerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminars
 
Central Composite Design
Central Composite DesignCentral Composite Design
Central Composite Design
 
Part03
Part03Part03
Part03
 
Agreement analysis
Agreement analysisAgreement analysis
Agreement analysis
 
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
A sensitivity analysis of contribution-based cooperative co-evolutionary algo...
 
Revisiting the Calibration of Modern Neural Networks
Revisiting the Calibration of Modern Neural NetworksRevisiting the Calibration of Modern Neural Networks
Revisiting the Calibration of Modern Neural Networks
 
ARIMA Model.ppt
ARIMA Model.pptARIMA Model.ppt
ARIMA Model.ppt
 
ARIMA Model for analysis of time series data.ppt
ARIMA Model for analysis of time series data.pptARIMA Model for analysis of time series data.ppt
ARIMA Model for analysis of time series data.ppt
 
ARIMA Model.ppt
ARIMA Model.pptARIMA Model.ppt
ARIMA Model.ppt
 
Ch06 alignment
Ch06 alignmentCh06 alignment
Ch06 alignment
 
Design of experiments(
Design of experiments(Design of experiments(
Design of experiments(
 
Biostatistics ppt.pptx
Biostatistics ppt.pptxBiostatistics ppt.pptx
Biostatistics ppt.pptx
 
Bounded Model Checking
Bounded Model CheckingBounded Model Checking
Bounded Model Checking
 
Reporting.ppt
Reporting.pptReporting.ppt
Reporting.ppt
 
Bioequivalence experimental study design By Vishnu Datta M
Bioequivalence experimental study design By Vishnu Datta MBioequivalence experimental study design By Vishnu Datta M
Bioequivalence experimental study design By Vishnu Datta M
 

Mais de NIHR CLAHRC West Midlands

Alison Hipwell presentation at PPI Supervisory Committee meeting
Alison Hipwell presentation at PPI Supervisory Committee meetingAlison Hipwell presentation at PPI Supervisory Committee meeting
Alison Hipwell presentation at PPI Supervisory Committee meetingNIHR CLAHRC West Midlands
 
Rapid qualitative analysis: does it deliver? - Beck Taylor
Rapid qualitative analysis: does it deliver? - Beck TaylorRapid qualitative analysis: does it deliver? - Beck Taylor
Rapid qualitative analysis: does it deliver? - Beck TaylorNIHR CLAHRC West Midlands
 
Shout Out for Youth Mental Health feedback - John Ovretveit
Shout Out for Youth Mental Health feedback - John OvretveitShout Out for Youth Mental Health feedback - John Ovretveit
Shout Out for Youth Mental Health feedback - John OvretveitNIHR CLAHRC West Midlands
 
Towards a National Learning Health System - Aziz Sheikh
Towards a National Learning Health System - Aziz SheikhTowards a National Learning Health System - Aziz Sheikh
Towards a National Learning Health System - Aziz SheikhNIHR CLAHRC West Midlands
 
A century of health services development in Birmingham - Tim Jones
A century of health services development in Birmingham - Tim JonesA century of health services development in Birmingham - Tim Jones
A century of health services development in Birmingham - Tim JonesNIHR CLAHRC West Midlands
 
4 - Further info - Setting up research in the NHS: practical and ethical cons...
4 - Further info - Setting up research in the NHS: practical and ethical cons...4 - Further info - Setting up research in the NHS: practical and ethical cons...
4 - Further info - Setting up research in the NHS: practical and ethical cons...NIHR CLAHRC West Midlands
 
3 - UoB - Setting up research in the NHS: practical and ethical considerations
3 - UoB - Setting up research in the NHS: practical and ethical considerations3 - UoB - Setting up research in the NHS: practical and ethical considerations
3 - UoB - Setting up research in the NHS: practical and ethical considerationsNIHR CLAHRC West Midlands
 
2 CRN - Setting up research in the NHS: practical and ethical considerations
2 CRN - Setting up research in the NHS: practical and ethical considerations2 CRN - Setting up research in the NHS: practical and ethical considerations
2 CRN - Setting up research in the NHS: practical and ethical considerationsNIHR CLAHRC West Midlands
 
1 - HRA - Setting up research in the NHS: practical and ethical considerations
1 - HRA - Setting up research in the NHS: practical and ethical considerations1 - HRA - Setting up research in the NHS: practical and ethical considerations
1 - HRA - Setting up research in the NHS: practical and ethical considerationsNIHR CLAHRC West Midlands
 
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom Marshall
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom MarshallCLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom Marshall
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom MarshallNIHR CLAHRC West Midlands
 
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...NIHR CLAHRC West Midlands
 
Prof Tom Marshall - Introduction to the opportunties and support available to...
Prof Tom Marshall - Introduction to the opportunties and support available to...Prof Tom Marshall - Introduction to the opportunties and support available to...
Prof Tom Marshall - Introduction to the opportunties and support available to...NIHR CLAHRC West Midlands
 
The last fellowship... the next fellowship application
The last fellowship... the next fellowship applicationThe last fellowship... the next fellowship application
The last fellowship... the next fellowship applicationNIHR CLAHRC West Midlands
 
Rapid qualitative analysis vs the 'traditional approach': early findings and ...
Rapid qualitative analysis vs the 'traditional approach': early findings and ...Rapid qualitative analysis vs the 'traditional approach': early findings and ...
Rapid qualitative analysis vs the 'traditional approach': early findings and ...NIHR CLAHRC West Midlands
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_dt
Clahrc ps cmeeting_21st_sept2015_spacer_project_dtClahrc ps cmeeting_21st_sept2015_spacer_project_dt
Clahrc ps cmeeting_21st_sept2015_spacer_project_dtNIHR CLAHRC West Midlands
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_cc
Clahrc ps cmeeting_21st_sept2015_spacer_project_ccClahrc ps cmeeting_21st_sept2015_spacer_project_cc
Clahrc ps cmeeting_21st_sept2015_spacer_project_ccNIHR CLAHRC West Midlands
 
Preventing Obesity in Pregnancy Study (POPS 2)
Preventing Obesity in Pregnancy Study (POPS 2) Preventing Obesity in Pregnancy Study (POPS 2)
Preventing Obesity in Pregnancy Study (POPS 2) NIHR CLAHRC West Midlands
 
Where's WALY - Aileen Clarke and Sian Taylor-Phillips
Where's WALY - Aileen Clarke and Sian Taylor-PhillipsWhere's WALY - Aileen Clarke and Sian Taylor-Phillips
Where's WALY - Aileen Clarke and Sian Taylor-PhillipsNIHR CLAHRC West Midlands
 
Publication bias widening the research lens - russell mannion
Publication bias widening the research lens - russell mannionPublication bias widening the research lens - russell mannion
Publication bias widening the research lens - russell mannionNIHR CLAHRC West Midlands
 

Mais de NIHR CLAHRC West Midlands (20)

Emma Healey presentation on ENHANCE Project
Emma Healey presentation on ENHANCE ProjectEmma Healey presentation on ENHANCE Project
Emma Healey presentation on ENHANCE Project
 
Alison Hipwell presentation at PPI Supervisory Committee meeting
Alison Hipwell presentation at PPI Supervisory Committee meetingAlison Hipwell presentation at PPI Supervisory Committee meeting
Alison Hipwell presentation at PPI Supervisory Committee meeting
 
Rapid qualitative analysis: does it deliver? - Beck Taylor
Rapid qualitative analysis: does it deliver? - Beck TaylorRapid qualitative analysis: does it deliver? - Beck Taylor
Rapid qualitative analysis: does it deliver? - Beck Taylor
 
Shout Out for Youth Mental Health feedback - John Ovretveit
Shout Out for Youth Mental Health feedback - John OvretveitShout Out for Youth Mental Health feedback - John Ovretveit
Shout Out for Youth Mental Health feedback - John Ovretveit
 
Towards a National Learning Health System - Aziz Sheikh
Towards a National Learning Health System - Aziz SheikhTowards a National Learning Health System - Aziz Sheikh
Towards a National Learning Health System - Aziz Sheikh
 
A century of health services development in Birmingham - Tim Jones
A century of health services development in Birmingham - Tim JonesA century of health services development in Birmingham - Tim Jones
A century of health services development in Birmingham - Tim Jones
 
4 - Further info - Setting up research in the NHS: practical and ethical cons...
4 - Further info - Setting up research in the NHS: practical and ethical cons...4 - Further info - Setting up research in the NHS: practical and ethical cons...
4 - Further info - Setting up research in the NHS: practical and ethical cons...
 
3 - UoB - Setting up research in the NHS: practical and ethical considerations
3 - UoB - Setting up research in the NHS: practical and ethical considerations3 - UoB - Setting up research in the NHS: practical and ethical considerations
3 - UoB - Setting up research in the NHS: practical and ethical considerations
 
2 CRN - Setting up research in the NHS: practical and ethical considerations
2 CRN - Setting up research in the NHS: practical and ethical considerations2 CRN - Setting up research in the NHS: practical and ethical considerations
2 CRN - Setting up research in the NHS: practical and ethical considerations
 
1 - HRA - Setting up research in the NHS: practical and ethical considerations
1 - HRA - Setting up research in the NHS: practical and ethical considerations1 - HRA - Setting up research in the NHS: practical and ethical considerations
1 - HRA - Setting up research in the NHS: practical and ethical considerations
 
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom Marshall
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom MarshallCLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom Marshall
CLAHARC WM Capacity Development Strategy - Nathalie Maillard and Tom Marshall
 
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...
Feedback from 'speed dating' - Postgrad / Early Career Researcher event 19th ...
 
Prof Tom Marshall - Introduction to the opportunties and support available to...
Prof Tom Marshall - Introduction to the opportunties and support available to...Prof Tom Marshall - Introduction to the opportunties and support available to...
Prof Tom Marshall - Introduction to the opportunties and support available to...
 
The last fellowship... the next fellowship application
The last fellowship... the next fellowship applicationThe last fellowship... the next fellowship application
The last fellowship... the next fellowship application
 
Rapid qualitative analysis vs the 'traditional approach': early findings and ...
Rapid qualitative analysis vs the 'traditional approach': early findings and ...Rapid qualitative analysis vs the 'traditional approach': early findings and ...
Rapid qualitative analysis vs the 'traditional approach': early findings and ...
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_dt
Clahrc ps cmeeting_21st_sept2015_spacer_project_dtClahrc ps cmeeting_21st_sept2015_spacer_project_dt
Clahrc ps cmeeting_21st_sept2015_spacer_project_dt
 
Clahrc ps cmeeting_21st_sept2015_spacer_project_cc
Clahrc ps cmeeting_21st_sept2015_spacer_project_ccClahrc ps cmeeting_21st_sept2015_spacer_project_cc
Clahrc ps cmeeting_21st_sept2015_spacer_project_cc
 
Preventing Obesity in Pregnancy Study (POPS 2)
Preventing Obesity in Pregnancy Study (POPS 2) Preventing Obesity in Pregnancy Study (POPS 2)
Preventing Obesity in Pregnancy Study (POPS 2)
 
Where's WALY - Aileen Clarke and Sian Taylor-Phillips
Where's WALY - Aileen Clarke and Sian Taylor-PhillipsWhere's WALY - Aileen Clarke and Sian Taylor-Phillips
Where's WALY - Aileen Clarke and Sian Taylor-Phillips
 
Publication bias widening the research lens - russell mannion
Publication bias widening the research lens - russell mannionPublication bias widening the research lens - russell mannion
Publication bias widening the research lens - russell mannion
 

Último

Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...
Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...
Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...Sheetaleventcompany
 
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetneemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...Ahmedabad Call Girls
 
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetErnakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Chandigarh
 
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...chandigarhentertainm
 
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetnagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...Ahmedabad Call Girls
 
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetpalanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetBihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In ChandigarhSheetaleventcompany
 
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetraisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetMathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Top 20 Famous Indian Female Pornstars Name List 2024
Top 20 Famous Indian Female Pornstars Name List 2024Top 20 Famous Indian Female Pornstars Name List 2024
Top 20 Famous Indian Female Pornstars Name List 2024Sheetaleventcompany
 
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetMangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetbhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girl
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girlKolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girl
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girlonly4webmaster01
 
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetThrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetThoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetCall Girls Service
 
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real Meet
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real MeetVip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real Meet
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real MeetAhmedabad Call Girls
 

Último (20)

Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...
Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...
Premium Call Girls Bangalore {7304373326} ❤️VVIP POOJA Call Girls in Bangalor...
 
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetneemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
neemuch Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...
(Deeksha) 💓 9920725232 💓High Profile Call Girls Navi Mumbai You Can Get The S...
 
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetErnakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Ernakulam Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...
❤️Call girls in Jalandhar ☎️9876848877☎️ Call Girl service in Jalandhar☎️ Jal...
 
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetnagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
nagpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...
Independent Call Girls Hyderabad 💋 9352988975 💋 Genuine WhatsApp Number for R...
 
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetpalanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
palanpur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetBihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Bihar Sharif Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh
💚 Punjabi Call Girls In Chandigarh 💯Lucky 🔝8868886958🔝Call Girl In Chandigarh
 
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetraisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
raisen Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetMathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mathura Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Top 20 Famous Indian Female Pornstars Name List 2024
Top 20 Famous Indian Female Pornstars Name List 2024Top 20 Famous Indian Female Pornstars Name List 2024
Top 20 Famous Indian Female Pornstars Name List 2024
 
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetMangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Mangalore Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetbhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
bhopal Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girl
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girlKolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girl
Kolkata Call Girls Miss Inaaya ❤️ at @30% discount Everyday Call girl
 
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetThrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thrissur Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real MeetThoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
Thoothukudi Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meetooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
ooty Call Girls 👙 6297143586 👙 Genuine WhatsApp Number for Real Meet
 
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real Meet
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real MeetVip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real Meet
Vip Call Girls Makarba 👙 6367187148 👙 Genuine WhatsApp Number for Real Meet
 

Design choices for cluster randomised trials - Alan Girling

  • 1. Design Choices for Cluster- Randomised Trials Alan Girling University of Birmingham, UK A.J.Girling@bham.ac.uk CLAHRC Scientific Advisory Group Birmingham, June 2015
  • 2. … a Statistical viewpoint • Ethical and Logistical concerns are suspended • RCTs and Parallel Cluster trials are well- understood; ‘Stepped Designs’ less so. Two questions: 1. How does the Stepped Wedge Design perform? …but The Genie is out of the bottle! 2. What about alternative stepped designs? Does it have to be a “Wedge”?
  • 3. Some Alternative Designs: 8 months 4 groups of Clusters (“Arms”) Arms 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Months  Arms 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Months 
  • 4. Assumptions • Study of fixed duration (8 months) • Constant recruitment rate in each cluster • Continuous Outcome – Additive treatment effect – Cross-sectional observations with constant ICC =  • Cross-over in one direction only (i.e. Treated to Control prohibited) • The analysis allows for a secular trend (“time effect”) – This has been questioned; but if time effects are ignored, the ‘best’ statistical design involves simple before-and-after studies in each cluster (i.e. not good at all!) Goal: To compare statistical performance of different designs under these assumptions, especially the Stepped-Wedge
  • 5. 1. ‘Precision Factor Plots’ for Comparing Designs
  • 6. Clusters 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Months  1. Simple Parallel Design Two groups of clusters. Treatment implemented in one group only, in month 1. Treatment Effect Estimate = (Mean difference between two groups over all months) Two Simple Candidate Designs
  • 7. Clusters 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Months  1. Parallel Study with (multiple) baseline controls “Controlled Before-and-After Design” Two groups of clusters. Treatment implemented in one group only, in month 5. Treatment Effect Estimate = (Mean difference between two groups in months 5 – 8) minus r x (Mean difference between two groups in months 1 – 4) (r is a correlation coefficient “derived from the ICC”)
  • 8. Performance measured by Precision of the effect estimate: Precision = 1/(Sampling Variance) = 1/(Standard Error)2 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0                R N 4 1 2 1 14 2     R N        1 14 2  Precision = Where R is the “Cluster-Mean Correlation” … and is the same for both designs    11   m m R (m = number of observations per cluster,  = ICC) CBA Parallel
  • 9. Relative Efficiency of designs – by comparing straight lines on a Precision-Factor plot CBA design is better (“more efficient”) if R > 2/3. Otherwise the Parallel design is better. Parallel CBA
  • 10. 2. The Cluster-Mean Correlation (CMC)
  • 11. Cluster-Mean Correlation (CMC) • Relative efficiency of different designs depends on cluster-size (m) and ICC (), but only through the CMC (R) • The CMC (R) = proportion of the variance of the average observation in a cluster that is attributable to differences between clusters • (The ICC () = proportion of variance of a single observation attributable to differences between clusters) • CMC can be large (close to 1) even if ICC is small       m m m m R     111
  • 12. Cluster-Mean Correlation: relation with cluster size • CMC can be large (close to 1) even if ICC is small • So CBA can be more efficient than a parallel design for reasonable values of the ICC, if the clusters are large enough      11 m R R
  • 13. 3. Some other designs
  • 14. 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Controlled Before & After Before & After + Parallel R 4 1 2 1  R 2 1 4 3  0 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 1 R 32 9 16 9  0 1 1 1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 1 R 16 5 8 5  + Baseline & Full Implementation Stepped Wedge Precision Factors
  • 15.
  • 16. • Stepped-Wedge is best when the CMC is close to 1 • Parallel Design is best when the CMC is close to 0 – but risky if ICC is uncertain • Before & After/Parallel mixture is a possible compromise Parallel Stepped-Wedge B & A/Parallel
  • 17. 4. Is there a ‘Best’ Design?
  • 18. • If Cross-over from Treatment to Control is permitted, the Bi-directional Cross-Over (BCO) design has Precision Factor = 1 at every R, better than any other design. 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 R01 Precision Factor = • But this design is usually not feasible! • If ‘reverse cross-over’ is disallowed, the Precision Factor cannot exceed 2 3 1 1 RR 
  • 19. 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 • Available design performance limited by choosing only 4 groups of clusters Some (nearly) ‘Best’ designs with 4 groups of clusters Stepped- Wedge Parallel Region Prohibited by Irreversibility of Intervention
  • 20. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Some (nearly) ‘Best’ designs with 8 groups of clusters Stepped- Wedge Parallel
  • 21. ‘Best’ Design for large studies: This is a mixture of Parallel and Stepped-Wedge clusters. 100R% Stepped-Wedge Clusters 100(1 – R)% Parallel Clusters • Includes Parallel and Stepped-Wedge designs as special cases • When R is close to 1 the Stepped-Wedge is the best possible design
  • 22. 5. Conclusions • Efficiency of different cluster designs depends on the ICC () and the Cluster-size (m) but only through the CMC (R). • Precision Factor plots are useful for comparing designs – Comparisons are linear in R • The Stepped-Wedge Design is most advantageous when R is close to 1. – In studies with large clusters this can arise even if the ICC is relatively small. • The theoretically ‘best’ design choice is sensitive to R, and combines Parallel with Stepped-Wedge clusters
  • 23. Limitations • Continuous data, simple mixed model – Natural starting point – exact results are possible – ?Applies to Binary observations through large- sample approximations • Extension to cohort designs through nested subject effects is straightforward and gives essentially the same answers